Integrated geophysical, hydrogeochemical and artificial intelligence techniques for groundwater study in the Langat Basin, Malaysia / Mahmoud Khaki

Mahmoud, Khaki (2014) Integrated geophysical, hydrogeochemical and artificial intelligence techniques for groundwater study in the Langat Basin, Malaysia / Mahmoud Khaki. PhD thesis, University of Malaya.

PDF (Full Text)
Download (6Mb) | Preview


    Geophysical, hydrogeochemical and artificial intelligence techniques were used to study the groundwater characteristics and their associated problems in Langat Basin, Malaysia. Resistivity surveys and geochemical analyses were used to delineate regions of Langat Basin that are contaminated by brackish water. Hydrogeochemical data of groundwater samples collected from seventeen wells from 2008 to 2013 were analysed. Ninety eight geoelectrical resistivity survey measurements were conducted to obtain subsurface resistivity data. The Wenner array was selected because of its sensitivity in detecting vertical changes in subsurface resistivity. The resistivity imaging results show that the upper layer is usually clay and below this layer is an aquifer with various depths of 10 to 30 m, and the layer thickness changes from 10 to 45 m, respectively, from east to west across the study area. The depth to bedrock varies from 30 m up to 65 m. The results learned from the resistivity survey confirmed the pattern of a continuous structure of layers, as detected from the borehole and geological information. Chemical analyses show the total dissolved solid exceeds 1000 mg/L in the west and is less than 1000 mg/L in the east of study area. Furthermore, the results of the resistivity survey and those from the hydrogeochemical analyses show that the groundwater within the study area is a mixture of brackish and freshwater zones. A novel investigation in modelling of groundwater level and quality using Artificial Neural Networks (ANNs) and Adaptive neuro fuzzy inference systems (ANFIS) methods was developed in the study area. Water table modelling based on ANNs and ANFIS technology were developed to simulate the water table fluctuations based on the relationship between the variations of rainfall, humidity, evaporation, minimum and maximum temperature and water table depth. The mean square errors and correlation coefficient of the water table depth models for 84 months were between 0.0043 to 0.107 and 0.629 to 0.99 respectively for all models. Evaluating the results of the various kinds iii of models show the earned results of the ANFIS model are superior to those gained from ANNs in which they are both more precise and with less error. Furthermore, four common training functions; Gradient descent with momentum and adaptive learning rate back propagation, Levenberg-Marquardt algorithm, Resilient back propagation, Scaled conjugate gradient were compared for the modelling of groundwater level. These results confirm that, for all the networks the Levenberg-Marquardt algorithm is the most effective algorithm to model the groundwater level. This study also developed the potential of the ANFIS and ANN to simulate total dissolved solid (TDS) and electrical conductivity (EC) by employing the values of other existing water quality parameters from five sampling stations over six years from 2008 to 2013. A good agreement between simulate values and their respective measured values in the quality of the groundwater were found. TDS and EC values predicted from the model accompanying obtained result present increasing in concentration continually in the future whereas for well no 3 (the nearest well to costal line) reaches at 32625.51 mg/L and 69501.76 μS/cm in 2025, respectively.

    Item Type: Thesis (PhD)
    Additional Information: Thesis (Ph.D.) -Faculty of science, University of Malaya, 2014.
    Uncontrolled Keywords: Intelligence techniques for groundwater study
    Subjects: Q Science > Q Science (General)
    Divisions: Faculty of Science
    Depositing User: Miss Dashini Harikrishnan
    Date Deposited: 12 Mar 2015 12:48
    Last Modified: 12 Mar 2015 12:48

    Actions (For repository staff only : Login required)

    View Item